Seminario Aleatorio

Sesión 373

Speeding up Inference on genetic trees and graphs

Felipe Medina Aguayo

Abstract

Sequencing DNA samples has become cheaper and faster than ever, as a consequence there is an enormous amount of data available to analyse. Many models in genetics involve the coalescent tree, a mathematical construction that has both discrete and continuous features. Performing inference on these trees is very challenging due to their complex nature, however Bayesian Statistics and advanced Monte Carlo methods provide some solutions. For the first part of the talk I will be discussing sequential Monte Carlo methodology for exploring tree spaces, which has some clear computational advantages over Markov Chain Monte Carlo (MCMC). For the second part I will introduce the concept of recombination in bacteria that gives rise to Ancestral Recombination Graphs (ARGs), a generalisation of coalescent trees. Inference on ARGs is typically performed via a reversible-jump MCMC which may perform poorly; however, as presented here, recent developments in Bayesian computation provide ways to improve existing methods and code that may lead to faster inference.

Viernes 01 de octubre de 2021, 13:00 hora de CDMX

https://itam.zoom.us/j/91443909666?pwd=c2JTNzVnSEJEN25IQXhqM1RVeEtXZz09

ID de reunión: 914 4390 9666

Código de acceso: 336509

El Seminario Aleatorio del Departamento de Estadística del ITAM está destinado tanto a profesores como a estudiantes, por lo que se agradece a los profesores que colaboren invitando a sus alumnos a estas sesiones.